2022: OpenAI, Stability.AI, Deepmind
Based on twtr buzz, search trends, etc.
Resolves ~50/30/20;
1,000
3.00@HanchiSun These markets seems to be about the impact the companies have had rather than raw technical sophistication. So, being very impressive but worse in every way than OpenAI isn't worth many points.
I've been nominated to resolve this market with a best guess as to what Gigacasting would have wanted (may he rest in peace).
This is judged based on the levels of buzz and hype each lab received during the year: Google Trends plus the vibes on my twitter timeline.
#1 - OpenAI again. After taking the world by storm in 2022, ChatGPT continued storming the world some more in 2023. GPT-4, DALL-E 3... The Sam Altman firing was so huge on Twitter that some people even imagined he would be named Time's Person of the Year.
#2 - Midjourney. This year Midjourney overtook StableDiffusion pretty decisively. I think this year's top-tier MJ outputs helped drive the fear and backlash among artists, which has been another one of the big AI narratives this year. Second place in search trends.
#3 - TIE: Facebook & Mistral. The two major contributors to the world of open-weights models. I couldn't pick between them because although the llamas have been workhorses throughout the year, they haven't generated much buzz while Mistral made a big splash with their EOY releases.
DISHONORABLE MENTIONS:
#4 - Anthropic. So far Anthropic seems to be famous for being "the lab that's almost as good as OpenAI" rather than anything any of their models have actually done.
#5 - Google/Deepmind. Gemini release was a flop on twitter. I don't know if Deepmind has done anything this year.
@Gigacasting feel free to request a re-resolve if you have your own take.
Alright, I'm off! See you all next year in https://manifold.markets/Gigacasting/what-will-be-the-top3-ai-labs-in-20-ff578e9dd7a2
@MichaelWheatley Nice, very gracefully done and seems entirely consistent with what I would have expected Gigacasting to have done
@MichaelWheatley How about these?
https://manifold.markets/Gigacasting/what-will-be-the-top3-ai-tools-in-2
https://manifold.markets/Gigacasting/ai-will-someone-train-a-100m-model
https://manifold.markets/Gigacasting/ai-leaderboard-ranking-aoc-2023
https://manifold.markets/Gigacasting/what-will-be-the-next-ai-breakthrou
That would resolve all of Gigacasting closed markets if those got resolved, I have no idea after searching which way those 4 go!!!!
I'll weigh in - Clearly the top labs are amongst:
OpenAI, Anthropic, Mistral, Deepmind, Perplexity, Google, and Meta this year.
However, I don't know whether this market is a popularity contest or a meaningful contribution contest.
Character.ai was the lab for 2022, not for 2023. I don't know anything notable they did for AI that they'd not done in 2022
Stability was the buzz for 2022, not 2023.
Meta and Google fall behind OpenAI, Anthropic and Mistral, but they did make impressive progress. Especially Google. I feel like their work in alignment and on fundamentals was strong this year.
Perplexity was good but not as much as OpenAI, Anthropic, and Mistral.
Mistral started in April 2023 with a banger start and has been the buzz of the town since.
Meta had Llama, llama2 and these are the go to models for 2023 open source other than mistral, all year.
Hence, my ranking would be, as an uninvested party in this market:
OpenAI -> Mistral -> Meta ~= Anthropic > Deepmind > Google > Midjourney > Perplexity
@firstuserhere However, if this was a popularity contest, then god knows how you'll resolve this. Probably google search or something, but that's not a very meaningful market
@firstuserhere Note that 'Google Deepmind' should be considered as one entity (70% to DeepMind, 30% to Google), assuming we're following Gigacasting's determination following the merger.
@SirCryptomind The weird thing is that if we're going by search trends, I'm not sure why Midjourney wasn't top 3 in 2022:
It was more popular than both stability ai and deepmind in trends.
Based on my anecdotal recollection of "twitter buzz", which is the other criteria in the description, I'd say that OpenAI and Deepmind were #1 and #2 in 2023. I have no idea how to fairly choose #3.
@SirCryptomind Well if FUH said that Mistral and Meta were the most technically impressive labs after OAI, I feel like at least one of those should be in the top 3?
I think if I was trying to guess how Gigacasting would resolve this, I'd say
1) OAI
2) Deepmind
3) Mistral
@Joshua I think @MichaelWheatley was mentioning mistral for this market too.
If we're going as gigacasting would, he'd rank his favourite company meta 1st, anything open source next, and openai probably 3rd lol.
@firstuserhere loooool
Maybe we say we're appointing Michael as the final resolver and this goes as he suggests?
@Joshua That's two likes from Cryptomind and FUH, @MichaelWheatley will you answer the call of duty?
@Joshua I'm personally not too bothered either way, but the spirit of these markets were that they would resolve according to Gigacasting's strong (but subjective) convictions, and it feels a bit unnatural to have a mod resolve them
@finn The only issue would be that if someone determines @Gigacasting is an abandoned account, which this market could probably N/A which is unfair to the 165 predictors in the market. I think we are trying to avoid a chance of a N/A here since it has been known this creator has not been active. I appreciate your input very much!